📌 TOPINDIATOURS Breaking ai: Researchers from PSU and Duke introduce “Multi-Agent

Share My Research is Synced’s column that welcomes scholars to share their own research breakthroughs with over 2M global AI enthusiasts. Beyond technological advances, Share My Research also calls for interesting stories behind the research and exciting research ideas. 

Meet the author
Institutions: Penn State University, Duke University, Google DeepMind, University of Washington, Meta, Nanyang Technological University, and Oregon State University. The co-first authors are Shaokun Zhang of Penn State University and Ming Yin of Duke University.

In recent years, LLM Multi-Agent systems have garnered widespread attention for their collaborative approach to solving complex problems. However, it’s a common scenario for these systems to fail at a task despite a flurry of activity. This leaves developers with a critical question: which agent, at what point, was responsible for the failure? Sifting through vast interaction logs to pinpoint the root cause feels like finding a needle in a haystack—a time-consuming and labor-intensive effort.
 
This is a familiar frustration for developers. In increasingly complex Multi-Agent systems, failures are not only common but also incredibly difficult to diagnose due to the autonomous nature of agent collaboration and long information chains. Without a way to quickly identify the source of a failure, system iteration and optimization grind to a halt.
 
To address this challenge, researchers from Penn State University and Duke University, in collaboration with institutions including Google DeepMind, have introduced the novel research problem of “Automated Failure Attribution.” They have constructed the first benchmark dataset for this task, Who&When, and have developed and evaluated several automated attribution methods. This work not only highlights the complexity of the task but also paves a new path toward enhancing the reliability of LLM Multi-Agent systems.
The paper has been accepted as a Spotlight presentation at the top-tier machine learning conference, ICML 2025, and the code and dataset are now fully open-source.

Paper：https://arxiv.org/pdf/2505.00212
Code：https://github.com/mingyin1/Agents_Failure_Attribution
Dataset：https://huggingface.co/datasets/Kevin355/Who_and_When
 
 
Research Background and Challenges
LLM-driven Multi-Agent systems have demonstrated immense potential across many domains. However, these systems are fragile; errors by a single agent, misunderstandings between agents, or mistakes in information transmission can lead to the failure of the entire task.

Currently, when a system fails, developers are often left with manual and inefficient methods for debugging:
Manual Log Archaeology : Developers must manually review lengthy interaction logs to find the source of the problem.
Reliance on Expertise : The debugging process is highly dependent on the developer’s deep understanding of the system and the task at hand.
 
This “needle in a haystack” approach to debugging is not only inefficient but also severely hinders rapid system iteration and the improvement of system reliability. There is an urgent need for an automated, systematic method to pinpoint the cause of failures, effectively bridging the gap between “evaluation results” and “system improvement.”

Core Contributions
This paper makes several groundbreaking contributions to address the challenges above:
1. Defining a New Problem: The paper is the first to formalize “automated failure attribution” as a specific research task. This task is defined by identifying the

2. failure-responsible agent and the decisive error step that led to the task’s failure.

Constructing the First Benchmark Dataset: Who&When : This dataset includes a wide range of failure logs collected from 127 LLM Multi-Agent systems, which were either algorithmically generated or hand-crafted by experts to ensure realism and diversity. Each failure log is accompanied by fine-grained human annotations for:
Who: The agent responsible for the failure.
When: The specific interaction step where the decisive error occurred.
Why: A natural language explanation of the cause of the failure.

3. Exploring Initial “Automated Attribution” Methods : Using the Who&When dataset, the paper designs and assesses three distinct methods for automated failure attribution:
All-at-Once: This method provides the LLM with the user query and the complete failure log, asking it to identify the responsible agent and the decisive error step in a single pass. While cost-effective, it may struggle to pinpoint precise errors in long contexts.
Step-by-Step: This approach mimics manual debugging by having the LLM review the interaction log sequentially, making a judgment at each step until the error is found. It is more precise at locating the error step but incurs higher costs and risks accumulating errors.
Binary Search: A compromise between the first two methods, this strategy repeatedly divides the log in half, using the LLM to determine which segment contains the error. It then recursively searches the identified segment, offering a balance of cost and performance.
 
Experimental Results and Key Findings

Experiments were conducted in two settings: one where the LLM knows the ground truth answer to the problem the Multi-Agent system is trying to solve (With Ground Truth) and one where it does not (Without Ground Truth). The primary model used was GPT-4o, though other models were also tested. The systematic evaluation of these methods on the Who&When dataset yielded several important insights:

• A Long Way to Go: Current methods are far from perfect. Even the best-performing single method achieved an accuracy of only about 53.5% in identifying the responsible agent and a mere 14.2% in pinpointing the exact error step. Some methods performed even worse than random guessing, underscoring the difficulty of the task.
• No “All-in-One” Solution: Different methods excel at different aspects of the problem. The All-at-Once method is better at identifying “Who,” while the Step-by-Step method is more effective at determining “When.” The Binary Search method provides a middle-ground performance.

• Hybrid Approaches Show Promise but at a High Cost: The researchers found that combining different methods, such as using the All-at-Once approach to identify a potential agent and then applying the Step-by-Step method to find the error, can improve overall performance. However, this comes with a significant increase in computational cost.

• State-of-the-Art Models Struggle: Surprisingly, even the most advanced reasoning m…

  Konten dipersingkat otomatis.

  🔗 Sumber: syncedreview.com

  ---

  📌 TOPINDIATOURS Eksklusif ai: You Might Want to Ditch Your AI Investments Now That

  We’re not in the business of making financial recommendations, but CNBC personality Jim Cramer says there’s no bubble coming for the AI sector — which, given his track record, should probably make any AI investors just a bit nervous.

  Now that Cramer — whose calls have historically underperformed so much that he inspired an entire phenomenon of people betting against him, called the “Inverse Cramer” effect — has chimed in on the topic, netizens are worried about an imminent collapse.

  “Oh god, it’s worse than we thought,” one Reddit user wrote in a tongue-in-cheek post. “The grim reaper of finance has weighed in, the collapse of the global financial system is imminent.”

  Cramer has made plenty of disastrous calls over decades. In February 2000, he proclaimed that internet-related companies “are the only ones worth owning right now” — right before the dot-com collapse. In 2012, he bet against stalwarts like HP, Netflix, both of which soared following his sell notice. Hes even been accused of playing a part in the 2008 financial crisis.

  His latest prognostication? That he doesn’t believe the immense spending on AI infrastructure has any parallels with the dot-com bubble from 25 years ago — yes, the same one he didn’t see coming back in the day, either.

  He’s breaking, of course, with experts who have drawn connections for quite some time now, with some arguing that the current AI bubble may be even worse than the market conditions leading up to the dot-com implosion of the early 2000s.

  The stakes are considerable, with investors warning that spending on AI has contributed more to the growth of the US economy so far this year than all of consumer spending combined, foreshadowing a rude awakening. The sheer amount of resources being poured into AI could even mean that a crash could take down the US economy with it, some have argued.

  As a result, Cramer has become the target of ridicule on social media.

  Jim Cramer says investors “should not worry about a US government shutdown” pic.twitter.com/AHTxj0zt9k

  — Inverse Cramer (@CramerTracker) September 30, 2025

  The trouble started when Cramer pushed back against dot-com bubble comparisons on Monday, arguing that Big Tech is simply too big to fail.

  “Speaking as an internet pioneer, what I see now is the polar opposite of what we were seeing 25 years ago,” he said.

  “When the dot-coms made bad investments, nearly all of them went under,” he added. “But, worst case scenario, if Google and Amazon and Meta make bad investments and take big losses, that’s just another day at the office.”

  Cramer argued that thanks to an enormous amount of available cash, tech giants could write off debt as needed — or pivot, if AI turns out to be a dead end.

  However, the analyst remained optimistic about AI tech becoming useful enough to justify enormous spending.

  Naturally, Cramer didn’t want to stick his neck out too much, arguing that he wasn’t willing to write off the “dot-com bomb scenario” entirely.

  “See, the skepticism keeps things in check,” he said. “If there weren’t such a negative bent to the story right now, everyone would be in this pool, and we’d all drown.”

  Whether others will agree with his line of thinking remains to be seen, especially as analysts see more and more red flags going up. Last week, AI chipmaker Nvidia announced a $100 billion investment in its own biggest customer, OpenAI, once again fueling fears over an AI bubble.

  Analysts called it out as an example of “circular financing,” with Nvidia propping up a company important to its own bottom line.

  “It’s kind of like having your parents co-sign on your first mortgage,” Bloomberg‘s Jay Goldberg argued, describing the deal as “bubble-like behavior.”

  “When times are good, this is going to make things better,” he added. “We’re going to grow faster; numbers are going to go up much faster.”

  “But when the cycle turns, and it will turn, it makes things worse on the downside,” he added.

  More on the AI bubble: The AI Bubble Bursting Would Actually Be Incredible for the Economy, Economist Says

  The post You Might Want to Ditch Your AI Investments Now That Jim Cramer Says No Bubble Is Coming appeared first on Futurism.

  🔗 Sumber: futurism.com

  ---